[opendtv] Re: Chroma Formats and Why Europe should choose 720P for HDTV

  • From: Craig Birkmaier <craig@xxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Sun, 19 Dec 2004 10:58:55 -0500

At 3:56 PM -0500 12/17/04, Bob Miller wrote:
>What is the trade off if we just compressed 720P more and pre filtered
>it some so that the bit rate would be the same as a 480P for broadcast
>anyway? You would lose no more in difficult frames than you do with 480P
>but with simpler scenes you would have true 720P. That is let the codec
>do the choosing instead of pre choosing 480P.

Modulation of resolution. The picture would grow softer or sharper 
based on information content. This is already happening with DBS, and 
to a lesser extent with digital cable.

If you want to deliver images of a given level of image quality, you 
must establish a level of quality that handles the peak demands 
without visible compression artifacts. If you do that, then 
everything else will be equally sharp, using less of the channel.

>Also by going from 720P to 480P, assuming the same compression, you are
>not saving as many bits in the broadcast as the reduced pixel count
>(408,960/921,600) would imply, 56%. The scenes are still just as
>complicated so with MPEG2 you might save only 25%  and with an advanced
>codec like MPEG4 you would save even less, maybe 15%.

There is some truth to this. Think of it this way.

As rasters sizes increase, what we are doing is raising the highest 
frequency of details that can be represented faithfully. At some 
rater size, say 240 x 160,  the highest frequency will be well below 
the highest frequency that the viewer can detect. As we increase the 
raster size while keeping the screen size constant, we will 
eventually reach the point where we saturate the human visual system.

It is clear that 480P does not saturate the visual system on displays 
that are larger than 40 inch diagonal and preferred viewing 
distances. As we move to 576P we add useful frequencies and get 
closer to saturating the visual system on big screens.

So when trying to choose the optimal emission format for a given 
application, we need to balance the information content and the 
encoding settings in such a way that we preserve the integrity of the 
source samples. I can clearly demonstrate 480P or 576P source that 
looks better than 720P source at a given bit rate. If we squeeze too 
hard, the higher resolution raster may contain less information than 
the lower resolution raster that benefits from oversampling and more 
overhead for peak bit rate demands.

>A compromise might also be in reducing from 1280 to 1024 or even 720 by
>720P. Or you could do both, compress 720P down more and also reduce from
>1280 to 1024 or 720 by 720P.

All legal options in MPEG, but not ATSC.

>Wouldn't these be better compromises than just going from 720P to 480P
>everything else being equal?

Forget 480P. Think 576P The difference is dramatic in both vertical 
and horizontal detail. But 720P may be the right choice if the 
application demands it and the channel can support it.

What is important is to forget about the numbers and focus on 
delivering high quality samples.

You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: